Template Adaptation for Face Verification and Identification
Nate Crosswhite, Jeffrey Byrne Systems、Technology Research (STR) {nate.crosswhite, Chris Stauffer Visionary Systems and Research (VSR) Omkar M. Parkhi, Qiong Cao, Andrew Zisserman University of Oxford {omkar, qiong,
Abstract
Face recognition performance evaluation has traditionally focused on one-to-one verification, popularized by the Labeled Faces in the Wild dataset [1] for imagery and the YouTubeFaces dataset [2] for videos. In contrast, the newly released IJB-A face recognition dataset [3] unifies evaluation of one-to-many face identification with one-to-one face verification over templates, or sets of imagery and videos for a subject. In this paper, we study the problem of template adaptation, a form of transfer learning to the set of media in a template. Extensive performance evaluations on IJB-A show a surprising result, that perhaps the simplest method of template adaptation, combining deep convolutional network features with template specific linear SVMs, outperforms the state-of-the-art by a wide margin. We study the effects of template size, negative set construction and classifier fusion on performance, then compare template adaptation to convolutional networks with metric learning, 2D and 3D alignment. Our unexpected conclusion is that these other methods, when combined with template adaptation, all achieve nearly the same top performance on IJB-A for template-based face verification and
中文速览
人脸识别领域长期以"一对一"配对验证为主要评测方式，但现实场景往往需要把同一人的多张图片和视频整合成一个"模板"再进行比对。针对这一更贴近实战的模板级识别问题，研究者提出了"模板自适应"（template adaptation）方法：先用在海量人脸上训练好的深度卷积网络提取特征，再在测试阶段针对每个模板单独训练一个线性SVM分类器，把该模板的媒体编码作为正例、大规模无关人脸特征作为负例，从而为每个新身份定制相似度打分函数。在最具挑战性的IJB-A公开数据集上进行大规模评测后，结果出人意料：这个看似简单的方案大幅超越了当时所有最先进方法，包括三元组损失度量学习、联合贝叶斯度量学习、3D姿态归一化等复杂技术；更令人惊讶的是，将上述复杂技术与模板自适应结合后，彼此最终性能趋于一致，说明模板自适应本身才是提升的核心驱动力。这项研究的重要意义在于，它揭示了当模板包含足够多媒体时，为每个身份量身定制一个简单的线性分类器就能显著提升识别性能，为工程实践提供了一条低成本、高收益的路径。
原文 arXiv:1603.03958;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1603.03958v3